A Novel Risk Score to Predict One-Year Mortality in Patients Undergoing Complex High-Risk Indicated Percutaneous Coronary Intervention (CHIP-PCI)

Abstract

Objective. To identify patients undergoing complex, high-risk indicated percutaneous coronary intervention (CHIP-PCI) and compare their outcomes with non-CHIP patients. We created a CHIP score to risk stratify these patients. Background. Risk stratification of PCI patients remains difficult because most scoring systems reflect hemodynamic instability and predict early mortality. Methods. CHIP-PCI was defined as any of the following: age >80 years; ejection fraction <30%; dialysis; prior bypass surgery; treatment of left main trunk; chronic total occlusion; or >2 lesions in >1 coronary artery. The primary endpoint was 1-year all-cause mortality. Logistic regression identified independent predictors of 1-year mortality and the odds ratios (ORs) for those predictors were used to create a CHIP score. Patients were then classified as low, intermediate, and high risk. Results. Among 4478 patients, a total of 1730 (38.6%) were CHIP. There were 85 deaths (2.2%) at 1 year (4.1% in CHIP patients and 1.0% in non-CHIP patients; P<.001). CHIP-PCI was an independent predictor of mortality (OR, 2.57; 955 confidence interval, 1.52-4.32; P<.001). Four CHIP criteria were independent predictors of mortality: age >80 years (3 points); dialysis (6 points); ejection fraction <30% (2 points); and number of lesions treated >2 (2 points). Accordingly, there were 2752 low-risk (score of 0), 889 intermediate-risk (score of 2-3), and 267 high-risk patients (score of 4-13). The 1-year mortality rates among these 3 groups were 1.24%, 2.47%, and 10.86%, respectively (P<.001). Conclusion. Compared with non-CHIP, CHIP-PCI is associated with increased risk of 1-year mortality, which is particularly evident among those fulfilling >1 CHIP criterion.  

J INVASIVE CARDIOL 2021;33(4):E253-E258. Epub 2021 February 4. doi:10.25270/jic/20.00515

Key words: CHIP-PCI, coronary disease, mortality

Coronary disease remains the major contributor to heart-disease related mortality despite significant advances in medical therapy.¹ Complex high-risk indicated percutaneous coronary intervention (CHIP-PCI) refers to clinically driven percutaneous revascularization of patients with extensive coronary artery disease (CAD). It requires a set of skills, personnel, equipment, and logistical support beyond those needed for regular PCI. The concept of CHIP was highlighted in a position paper from 2016.² The assignation of CHIP relied on demographic factors (age), comorbid conditions (advanced kidney disease, prior coronary artery bypass grafting [CABG], heart failure), or anatomical and procedural considerations (extent of CAD, treatment of left main trunk [LMT], or chronic total occlusion [CTO] lesions, use of mechanical cardiac support, or use of atherectomy devices). Yet, a clear definition for CHIP has not been agreed upon and data regarding the outcome of CHIP relative to more usual PCI or CABG are lacking. A recent compilation of data from the British PCI registry suggested that greater experience with CHIP may reduce morbidity, but not mortality.³

Thus, we set out to identify CHIP patients in our institutional PCI registry using a pragmatic definition and compare their baseline characteristics and outcomes with those undergoing non-CHIP PCI. We used these data to create a useful score for predicting outcomes in order to assist practitioners endeavoring to revascularize this high-risk population.

Methods

Patients undergoing PCI at our tertiary-care hospital are enrolled in a registry collecting data for mandatory reporting purposes to the State Department of Health and the National Cardiovascular Data Registry (NCDR), using proprietary software from the American College of Cardiology (NCDR versions 4.4 and 5.0). The registry receives annual waiver of patient consent for participation based on aggregate, de-identified reporting of data. We identified all PCI procedures from January 1, 2014 through December 31, 2018 (n = 6078). Data collection is based on site assessment of demographic and procedural characteristics with minimal independent auditing. Many characteristics, such as vessel diameter, lesion length, presence of bifurcation lesion, Thrombolysis in Myocardial Infarction (TIMI) flow grade, and others, are never audited. We excluded patients presenting with acute ST-segment elevation myocardial infarction (STEMI), those in cardiogenic shock (n = 476), or those who had repeated procedures during the study period (n = 1124). The primary study endpoint was mortality at 1 year, which was collected by searching the individual electronic medical record of each patient, or contacting the patient, family, or referring physician when necessary. 

Patients were characterized as CHIP if they had at least 1 of the following characteristics associated with higher mortality after PCI: age >80 years; left ventricular ejection fraction (LVEF) <30% before PCI; end-stage renal disease on dialysis (ESRD); prior CABG; treatment of LMT or CTO; or treatment of >2 lesions during the procedure in at least 2 coronary arteries.^4,5 Patients without any of these features were categorized as non-CHIP. We did not include atherectomy as a criterion for CHIP because of very low utilization (<5%).

Statistical analysis. Continuous variables are presented as medians with interquartile range (IQR) and were compared with Wilcoxon’s rank-sum tests. Categorical variables are presented as proportions and were compared with Chi-squared or Fisher’s exact tests.  Multivariable logistic regression model was performed to test the independent impact of CHIP or its defining components on 1-year mortality (as date of death was rarely available, Cox proportional hazards analysis could not be performed). The following covariates were considered for all models: sex; hypertension; diabetes; prior MI; peripheral artery disease; prior PCI; prior CABG; and multivessel CAD. Each model was tested for goodness-of-fit (Hosmer-Lemeshow) and area under receiver operating characteristic (ROC) curve. We then created a CHIP score using the integer of the odds ratio (OR) for all significant independent predictors of mortality (only variables included in CHIP definition) and classified patients according to low (score = 0), intermediate (score = 2 or 3, having 1 CHIP criterion except dialysis), or high score (score >3, fulfilling more than 1 criterion for CHIP and/or being on dialysis). The significance level was set at P≤.05. All analyses were performed with STATA, version 14.2 (STATA).

Results

After exclusion criteria were applied, there were 4478 unique patients undergoing PCI during the study period. At least 1 CHIP characteristic was present in 1730 patients (38.6%). The number of CHIP features present were: 1 in 1259 patients (72.8%), 2 in 380 patients (22.0%), 3 in 81 patients (4.7%), and 4 in 10 patients (0.6%). The demographics, comorbidities, procedural details, and outcomes of the 2 groups are shown in Table 1. As expected, the CHIP and non-CHIP patients are significantly different in almost all comparisons. Of note is the very high prevalence of diabetes mellitus (55% in both groups; P=.80).

Death at 1 year occurred in 85 patients (2.2%), significantly more in the CHIP group vs the non-CHIP group (4.1% vs 1.0%, respectively; P<.001). Only 4% of all deaths (n = 3) occurred before hospital discharge and none were intraprocedural. The independent predictors of death are shown in Table 2. The model had excellent calibration (parameter goodness of fit [PGOF], 0.63) and good discrimination (area under the curve [AUC], 0.78). If the components of CHIP are supplanted by the composite CHIP variable, the model remains highly predictive (PGOF, 0.51; AUC, 0.78). CHIP-PCI significantly increased risk of death vs non-CHIP PCI (OR, 2.57; 95% CI, 1.52-4.32; P<.001). Similarly, the number of CHIP features were predictive of mortality with increased risk for each additional characteristic (OR, 1.98; 95% CI, 1.55-2.54; P<.001). Mortality data were missing in 571 patients (230 CHIP patients and 341 non-CHIP patients).

Based on the logistic regression model, a CHIP score was created. Four criteria for CHIP were significant predictors of mortality (P<.05) and assigned point values: age >80 years (3 points); dialysis (6 points); ejection fraction <30% (2 points); and number of lesions treated >2 (2 points). Thus, the score varied from 0 to 13. The median score was 0 among the 341 non-CHIP patients and 3 among the 230 CHIP patients without known vital status, which is not different than the distribution among the patients with known vital status at 1 year. Accordingly, among patients with known vital status, there were 2752 low-risk (score of 0, of which 356 patients were CHIP), 889 intermediate-risk (score of 2-3, fulfilling 1 CHIP criterion except dialysis), and 267 high-risk patients (score of 4-13, fulfilling 2 or more CHIP criteria and/or being on dialysis). The 1-year mortality rates among these 3 groups were 1.24%, 2.47%, and 10.86%, respectively (P<.001) (Figure 1). CHIP patients with score of 0 (qualified for CHIP because of treatment of LMT or CTO lesion, or prior CABG) and non-CHIP patients had similar mortality (1.69% vs 1.17%, respectively; P=.41). 

As a sensitivity analysis, we performed bootstrap replication (n = 50) of the multivariable regression model for death without dropping patients without vital status. There were no significant changes in predictors of mortality or the derived CHIP score. 

Discussion

We evaluated a large cohort of patients without STEMI or cardiogenic shock undergoing PCI at a single tertiary institution and identified those satisfying criteria for CHIP-PCI based on characteristics associated with higher mortality after PCI. The main findings of our study can be summarized as follows: (1) Nearly 40% of patients have CHIP features and their 1-year all-cause mortality, after adjustment for all other important baseline characteristics, is 2.5-fold higher than non-CHIP patients. This confirms the designation of patients as CHIP, even though certain procedural characteristics (LMT or CTO treatment), or prior CABG, thought to confer higher risk, were not independent predictors of mortality; (2) patients with more CHIP features have higher mortality than those with fewer features; and (3) a CHIP score derived from this dataset effectively describes the mortality risk of PCI and discriminates between those with higher vs lower risk; 

CHIP-PCI has become a subspecialty in interventional cardiology because it addresses a population with extensive CAD in need of revascularization and that has many risk factors for procedural and long-term adverse events. The benefits of revascularization in addition to optimal medical therapy in this population have been questioned by some studies, but confirmed in a large network meta-analysis encompassing over 93,000 patients enrolled in 100 trials. CABG and PCI with second-generation drug-eluting stents had similar reduction in mortality compared with medical therapy alone (relative risk [RR], 0.80; 95% CI, 0.70-0.91 and RR, 0.75; 95% CI, 0.59-0.96, respectively).⁶ It is not the intent of this analysis to define whether PCI is indeed beneficial in addition to medical therapy, but rather to clarify the risk associated with certain demographic, clinical, and angiographic features when PCI is undertaken.

The typical CHIP-PCI patient has significant kidney function impairment, diabetes mellitus, and depressed LVEF, while presenting with complex lesions in multiple territories. The need for more expensive and diverse equipment, attention to radiation and contrast media utilization, skills to treat complications, and need for aggressive adjunctive medical therapy have all been well articulated by Kirtane et al.² The interaction between baseline comorbidity, complex anatomy with concomitant valvular disease and/or left and right ventricular dysfunction, and impaired hemodynamics may render many of these patients unsuitable for surgical intervention.⁷

Recently, Kinnaird et al³ evaluated CHIP-PCI in the British Cardiovascular National Society PCI registry (BCIS). The authors used similar criteria as we did (plus use of atherectomy devices) to identify CHIP patients between 2007 and 2014.As in our series, by 2014, 36.2% of cases were CHIP-PCI. The 1-year mortality rate was 4.8%, which is remarkably similar to the rate observed in the current study, and did not seem to vary significantly with operator CHIP volume (P=.64). Higher-volume operators, however, did have a lower incidence of periprocedural non-fatal complications, predominantly related to CTO-PCI and access-site major bleeding. 

In our analysis, we chose not to include mechanical cardiac support (MCS) as a CHIP-defining feature because of the inherent variability in its utilization and lack of robust data regarding improvement in mortality. For example, in PROTECT II,⁸ patients undergoing high-risk PCI (LMT or 3-vessel CAD with LVEF<30%) were randomized to intra-aortic balloon pump (IABP) or Impella 2.5 L. At 30 days, there were no significant differences in major adverse cardiac events (40.1% vs 35.1%, respectively; P=.23) despite superior hemodynamic support offered by Impella.⁸ The lack of a control arm without MCS renders this comparison difficult to interpret. Even in cardiogenic shock patients studied in the IABP-SHOCK II trial, IABP did not confer a survival benefit when compared with routine PCI. Mortality at 1 year was 52% vs 51%, respectively; P=.91.⁹ It is notable that MCS was used in only 43 patients (1%) in our cohort and all were CHIP-PCI because of other criteria. 

To our knowledge, this is the first CHIP-PCI specific score to stratify risk of patients undergoing PCI with respect to 12-month all-cause mortality. Other risk scores are predominantly focused on in-hospital or 30-day outcomes and are heavily influenced by hemodynamic instability and clinical presentation with STEMI. In an earlier cohort of 3165 patients, we compared the accuracy of the Mayo Clinic risk score (MCRS)¹⁰ and New York risk score (NYRS)¹¹ for predicting in-hospital mortality and found both to have good calibration and discrimination abilities, but did not offer any information on long-term outcomes.¹² As we excluded patients with STEMI or cardiogenic shock and nearly all deaths in this cohort occurred after hospital discharge, the utilization of these scores, or a similar one from the Society of Coronary Angiography and Intervention (SCAI), which predicts in-hospital death, acute kidney injury, or need for transfusion, is not helpful.¹³ 

The most widely used score for decision making and outcome prediction in patient with CAD is the SYNTAX (Synergy Between PCI With TAXUS and Cardiac Surgery) score, originally derived to standardize angiographic interpretation of complex CAD in patients randomized to PCI or CABG.¹⁴ It was subsequently expanded to include clinical characteristics as well (SYNTAX II score).¹⁵ When used to predict 4-year mortality, the SYNTAX score had very modest discrimination (c-statistic 0.57 in SYNTAX trial derivation set and 0.62 in validation from the DELTA registry). The SYNTAX II score was somewhat more precise, with c-statistics of 0.73 and 0.72, respectively. We compared the 2 scores in a previous analysis from our PCI registry (831 patients followed for 4 years) and found a c-statistic of 0.75 for the SYNTAX II score.¹⁶ All of these are substantially lower than the precision of the CHIP score described in this manuscript. Furthermore, the complexity and lack of consistency in calculating the SYNTAX score renders a less than ideal predictor of outcomes.

Our data suggest a 9-fold higher mortality at 1 year in patients with CHIP score of 4 or higher, which can be merely achieved by being on dialysis or having both reduced LVEF and advanced age, compared with non-CHIP patients or CHIP patients with LMT or CTO treatment, or prior CABG, and without other features of high risk. If validated in other cohorts, or prospectively tested in a registry, the CHIP score may modify our approach to PCI in patients with advanced comorbidity and age by quantifying the risk associated with the procedure and the possibility of deferring it with further intensification of medical therapy. Even patients with intermediate risk (score 2-3, fulfilling 1 CHIP criterion) have double the mortality of those with lower scores.

Strengths and limitations. Our analysis is based on consecutive patients treated at a large-volume tertiary institution and avoids the pitfalls of patient selection. Yet, as the practice of interventional cardiology has evolved and the indications for PCI in patients with stable CAD have been refined,^17-19 the selection of patients for PCI has changed as well. We used a very specific definition for CHIP, but probably sacrificed sensitivity in this process. For example, patients with lower degrees of kidney dysfunction were not considered CHIP and treatment of a single remaining vessel may also not have been classified as CHIP if LVEF was preserved and other CHIP elements were not present. Similarly, patients with lesser degree of left ventricular dysfunction were not categorized as CHIP. Despite this, nearly 40% of the entire cohort were CHIP patients, albeit with a majority fulfilling only 1 criterion. It was notable that diabetes mellitus was not an independent risk factor for death in our study, likely because of the very high prevalence of the disease (55%). As mentioned above, we did not include MCS use as a CHIP criterion because of the marked selection bias and lack of supporting evidence for their benefit. We were surprised to find a lack of association between LMT treatment and mortality. It is possible that the relatively low number of LMT procedures performed contributed to this observation. We recognize that these are patients treated in the course of regular practice and not all patients had clear evidence that viable myocardium is present in the distribution of the treated lesions. We did not have survival data for 13% of the cohort despite intensive attempts, and neither did we have cause of death. Bootstrap analyses were performed to mitigate this deficiency as much as possible. Finally, we divided the risk strata rather arbitrarily in order to have enough representation in each category, thus allocating patients with only 1 significant predictor of mortality to the intermediate-risk category, while those with 2 or more criteria (or dialysis by itself) were classified as high risk. 

Conclusion

We conclude that in patients with CAD undergoing PCI in the absence of STEMI or cardiogenic shock, the 1-year mortality is about 2%. CHIP-PCI, defined by the presence of age >80 years, LVEF <30% before PCI, dialysis, prior CABG, treatment of LMT or CTO, or >2 lesions in at least 2 coronary arteries, is associated with a 2.5-fold higher risk of death. We derived a CHIP score (0-13 points) that accurately discriminates risk, such that patients with intermediate (2-3) and high scores (>3) have 2-fold and 9-fold higher 1-year mortality rate, respectively, than those with a score of 0. The findings need to be replicated and validated in a larger dataset.

References

1. Virani SS, Alonso A, Benjamin EJ, et al; American Heart Association Council on Epidemiology and Prevention Statistics Committee and Stroke Statistics Subcommittee. Heart disease and stroke statistics-2020 update: a report from the American Heart Association. Circulation. 2020;141:e139-e596.

2. Kirtane AJ, Doshi D, Leon MB, et al. Treatment of higher-risk patients with an indication for revascularization. Evolution within the field of contemporary percutaneous coronary intervention. Circulation. 2016;134:422-431.

3. Kinnaird T, Gallagher S, Spratt JC, et al. Complex high-risk and indicated percutaneous coronary intervention for stable angina: does operator volume influence patient outcome? Am Heart J. 2020;222:15-25.

4. Wu C, Camacho FT, King SB 3rd, et al. Risk stratification for long-term mortality after percutaneous coronary intervention. Circ Cardiovasc Interv. 2014;7:80-87. Epub 2014 Jan 14.

5. Othman H, Seth M, Zein R, et al; BMC2 Investigators. Percutaneous coronary intervention for chronic total occlusion- the Michigan experience: insights from the BMC2 registry. JACC Cardiovasc Interv. 2020;13:1357-1368. Epub 2020 May 13.

6. Windecker S, Stortecky S, Stefanini GG, et al. Revascularisation versus medical treatment in patients with stable coronary artery disease: network meta-analysis. BMJ. 2014;348:g3859.

7. Bortnick AE, Epps KC, Selzer F, et al. Five-year follow-up of patients treated for coronary artery disease in the face of an increasing burden of co-morbidity and disease complexity (from the NHLBI dynamic registry). Am J Cardiol. 2014;113:573-579.

8. O'Neill WW, Kleiman NS, Moses J, et al. A prospective, randomized clinical trial of hemodynamic support with Impella 2.5 versus intra-aortic balloon pump in patients undergoing high-risk percutaneous coronary intervention: the PROTECT II study. Circulation. 2012;126:1717-1727. Epub 2012 Aug 30.

9. Thiele H, Zeymer U, Neumann F-J, et al. Intra-aortic balloon counterpulsation in acute myocardial infarction complicated by cardiogenic shock (IABP-SHOCK II): final 12 month results of a randomised, open-label trial. Lancet. 2013;382:1638-1645. Epub 2013 Sep 3.

10. Singh M, Lennon RJ, Holmes DR Jr, Bell MR, Rihal CS. Correlates of procedural complications and a simple integer risk score for percutaneous coronary intervention. J Am Coll Cardiol. 2002;40:387-393.

11. Wu C, Hannan EL, Walford G, et al. A risk score to predict in-hospital mortality for percutaneous coronary interventions. J Am Coll Cardiol. 2006;47:654-660.

12. Brener SJ, Colombo KD, Haq SA, Bose S, Sacchi TJ. Precision and accuracy of risk scores for in-hospital death after percutaneous coronary intervention in the current era. Catheter Cardiovasc Interv. 2010;75:153-157.

13. Gurm HS, Seth M, Kooiman J, Share D. A novel tool for reliable and accurate prediction of renal complications in patients undergoing percutaneous coronary intervention. J Am Coll Cardiol. 2013;61:2242-2248.

14. Serruys PW, Morice MC, Kappetein AP, et al. Percutaneous coronary intervention versus coronary-artery bypass grafting for severe coronary artery disease. N Engl J Med. 2009;360:961-972.

15. Farooq V, van Klaveren D, Steyerberg EW, et al. Anatomical and clinical characteristics to guide decision making between coronary artery bypass surgery and percutaneous coronary intervention for individual patients: development and validation of SYNTAX score II. Lancet. 2013;381:639-650.

16. Brener SJ, Alapati V, Chan D, et al. The SYNTAX II score predicts mortality at 4 years in patients undergoing percutaneous coronary intervention. J Invasive Cardiol. 2018;30:290-294.

17. Levine GN, Bates ER, Blankenship JC, et al. 2011 ACCF/AHA/SCAI guideline for percutaneous coronary intervention: executive summary: a report of the American College of Cardiology Foundation/American Heart Association Task Force on Practice Guidelines and the Society for Cardiovascular Angiography and Interventions. J Am Coll Cardiol. 2011;58:2550-2583.

18. Neumann FJ, Sousa-Uva M, Ahlsson A, et al; ESC Scientific Document Group. 2018 ESC/EACTS guidelines on myocardial revascularization. Eur Heart J. 2019;40:87-165.

19. Boden WE, O'Rourke RA, Teo KK, et al. Optimal medical therapy with or without PCI for stable coronary disease. N Engl J Med. 2007;356:1503-1516.

From the Division of Cardiology, New York Presbyterian Brooklyn Methodist Hospital, Brooklyn, New York. 

Disclosure: The authors have completed and returned the ICMJE Form for Disclosure of Potential Conflicts of Interest. The authors report no conflicts of interest regarding the content herein.

Manuscript accepted August 21, 2020.

Address for correspondence: Sorin J. Brener, MD, FACC, Professor of Medicine, Director, Cardiac Catheterization Laboratory, New York Methodist Hospital, 506 6th Street, KP-2, Brooklyn, NY 11215. Email: sjb9005@nyp.org